Gabriel Lopes
5.9K posts
🚨 WAR UPDATE:
The straight is closed.
The straight is totally open.
The straight is closed.
Iran has agreed to ceasefire terms.
The straight is more closed.
Iran won’t agree to ceasefire terms.
We opened the straight.
They closed it.
They’re about to agree.
-all within past 5min
English
@lucas_montano Yep, just --model claude-opus-4-5-20251101
Still their best model, they have been regressing for the past 6 months
It is eating considerably more tokens though
English
Armchair psychologists, what is Theo supposed to do here? He's clearly hurting and broken. He's been consistently depressed for so long, his emotional state is tied to the opinions of others, he changes his beliefs every other day, how does this guy heal?
English
Funny how AI has completely killed the JavaScript framework wars.
There is absolutely no reason to use anything apart from React.
English
@evertjr @AnthropicAI Might wanna go back one more and you'll be golden
English
/model claude-opus-4-6
4.7 é regressão em todas métricas possíveis. Primeiro grande fail da @AnthropicAI
Português
@HackingDave We are doing the same thing here, when the switch to 4.6 happened took us nearly 30 days to notice the regression, hallucinations that would pop as issues a few days later, went back to 4.5 ever since
Attempted the 4.7 for a day, not looking forward to mythos
English
For the enterprises using Claude, if you are using it for heavy enterprise type stuff - be extremely careful. It's introducing massive bugs, security issues, and code quality is way worse than Opus 4.5, substantially worse on both 4.6 and 4.7.
Our entire development team is shifting off of it. It's unusable at the moment aside from beautiful UI stuff, it's code quality is not something you can trust.
Still no word from Claude on why they mangled their models and didn't tell anyone - which is particularly alarming on every front.
I would recommend switching teams over to something like Cursor, Perplexity, or AWS Bedrock - as the frontier models continue to innovate (or regress) - having the ability for flexible model selection that doesn't disrupt development workflow will be insanely important for enterprise.
English
@RhysSullivan Its funny people think typewriters follow rules. you can set codex in a sandbox in read-only mode and it will write stuff and run commands anyway
It will sometimes stop and say hey I'm in sandbox mode, but will later just do whatever
English
@DanielvsBabylon 40y in the desert now in the substance, trust He will provide daily and it will be sufficient
English
They don’t tell you this but your walk with Jesus will be messy. You will break down, you will cry, you will get angry, you will be frustrated, you will deal with doubt, you will deal with unbelief, you’re gonna mess up, you’re gonna do things you later regret. But through all this Christ will not leave you, and you will be sanctified through your trials. So if you were expecting a sunshine and rainbows walk with Jesus you are sorely mistaken. The storms will still come, but you’ll have God with you every step of the way, and your joy will be full when you come to terms with the fact that this is sufficient.
English
@zackkanter K8s look like it was a specific infra solution for a specific problem they had and people thought it could be generalized but the setup is too painful and not worth the payout
English
@mert fr fr
I attempted some tuis like tabby warp etc
Back to raw terminal
English
@gazollajr @dionisiodev both are very similar, of all the languages I've ever had to use those were indeed the most comfortable
English
@dionisiodev Das linguagens de programação modernas,
acho agradável a sintaxe do Kotlin e Swift.
Português
Eu entendo o conceito e as vantagens do Rust, mas, olha que sintaxe feia se comparar com C
Português
@thekitze It's been close to a year, every week there is a release of the same thing, a new ui to use the same tool of almost a year ago
The actual tool upgrade is right now mostly regressions
English
for the first time I feel what anti ai ppl have been saying
we might've reached peak and plateaued
llms might be morons
English
@pmddomingos Yes, anthropics best move would be to get sonnet to be as close to opus 4.5, that would cut their costs by half
They cant make their expensive models any better anyway
English
A lot of what passes for progress in AI is just the application of ever-increasing amounts of compute. And as a user you don't care until you start paying the bill.
English
@VictorTaelin 4.5 peaked, 4.6 was a downgrade, this is shaping up to be another downgrade, soon local models will pass them by simply not updating anything
English
I don't think we're all hallucinating, there's something seriously wrong about 4.7. Just tried it on the same two prompt (what's the best GC approach for Bend). 4.7 simply lies a lot, ignores information right on its context, makes bad proposals. This is really weird?
English
@theo We have reached greater and greater diminishing returns for over a year
This is it for LLMs, which is pretty good btw
People forget they are language models, good at languaging, they are not thinking anything
English
"Clearly Claude Code isn't that bad, millions of people use it!"
This mindset is how we got stuck with Internet Explorer.
I'm gonna keep pushing for us to do better as an industry.
English
@theo So far nope, for the past 6 months I use claude as main driver and codex to verify/double check and run some analysis/audit in the background, it has been pretty consistent
So has opus 4.5 btw, except it hits the 200k tokens 3x faster it seems nowadays
English
Serious question. Has anyone ever noticed meaningful regressions in Codex/OpenAI models?
I feel like we talk about this a lot w/ Anthropic but I've never seen a similar discussion with OAI.
English
@benhylak If they just kept 4.5 as it was in nov 2025 they'd still be king
All they had to do was nothing
English
anthropic has lost the divine mandate.
opus 4.7 is a bad model.
English